[Sis-SCPS-INTEREST] FYI-Newly published results on SCPS over actual satellite links

[Lloyd Wood]

A timer issue that almost all TCP-in-space papers neglect is the initial
SYN/ACK exchange, how long it would take, and whether the sending
TCP would time out or resend the SYN - and if so, how may times a
resend might occur.

If the SYN/ACK exchange times out and the TCP sender gives up (and real
implementations will always give up eventually), no communication is
possible, and any tweaks to timers related to congestion control and
windowing are entirely moot, as the TCP connection never gets
established, much less starts sending data.

A lot of TCP simulation in the literature was done using ns's one-way
TCP, which originally didn't simulate SYN/ACK at all (ns's FullTcp was
more complete). This omission appears to have adversely affected the
literature on TCP - papers on TCP-Peach and TP-Planet neglect the SYN/ 
ACK
issue entirely, for example, and this omission follows on to papers
surveying this area of research. Any discussion of long delays should
consider the SYN/ACK and FIN/ACK timer issues. I'd suggest revisiting
your 'problem solved: long delay' column in Table II.

Some further discussion of this is in:
Lloyd Wood, Cathryn Peoples, Gerard Parr, Bryan Scotney, Adrian Moore,
'TCP's protocol radius: the distance where timers prevent  
communication,'
International Workshop on Space and Satellite Communications (IWSSC  
'07),
Salzburg, Austria, September 2007. http://dx.doi.org/10.1109/IWSSC.2007.4409409

ns was amended to not repeat SYN/ACK indefinitely and better match
implementations as a result of this work.

But this is a minor point. A more significant point is that TCP and its
assumptions are unsuitable for the long delays or loss!=congestion
or dedicated-single-flow environments of space links, and TCP simply
won't be implemented. A number of satellites are operating using IP -
but don't implement TCP onboard.


On 29 Jun 2009, at 18:07, Ruhai Wang wrote:
> You may also be interested in the following survey paper. Scott  
> Burleigh at
> JPL is highly acknowledged for significant discussion.
> ------------------
> R. Wang, T. Taleb, A. Jamalipour, and B. Sun, "Protocols for  
> reliable data
> transport in space Internet," IEEE Communications Surveys and  
> Tutorials,
> vol. 11, No. 2, Second Quarter 2009, pp. 21-32.
> http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=5039581&isnumber=5039578 
> .

DTN work: http://info.ee.surrey.ac.uk/Personal/L.Wood/saratoga/

<http://info.surrey.ac.uk/Personal/L.Wood/><L.Wood at surrey.ac.uk>